TECHNICAL FIELD OF THE INVENTION

[0001]
The present invention relates in general to data processing, and in particular to secure logging of data in a file for irrefutable administration.

[0002]
In particular, the present invention finds an advantageous, but not exclusive, application in contexts involving system and network administrators and administered systems and networks, to which the following description will refer without this implying any loss of generality.
BACKGROUND ART

[0003]
As is known, there are contexts where a need exits to check and verify operations performed by an entity, e.g. an administrator, on another entity, e.g., a computer system. For example, in industry outsourcing is an adopted practice and therefore a need exits to control the operations performed by the personnel of the external company to which the job is farmed out and at the same time to guarantee the privacy of the personnel, without however giving up the possibility of verifying and linking, in case of necessity, the operations with the people who performed them.

[0004]
In a context involving system and network administrators and administered systems and networks, the aforementioned need is generally met by generating a socalled “log file”, which M. Ruffin, A survey of logging uses, Tech. Rep., Dept. of Computer Science, University of Glasgow, Glasgow, Scotland, February 1995. defines as a “plain file where data are stored sequentially as they arrive, by appending them to the end of the file. When a problem arises in the system (e.g. a fault or an intrusion), the log file is reread to find its source and/or to correct its consequences”.

[0005]
In many computer applications, sensitive information such as log files must be kept on an untrusted machine. Such information must be protected against attackers, as well as against partially trusted entities to be given partial, but not total, access to the stored information.

[0006]
US 2003/0212899 discloses a method, an apparatus, and computer instructions for protecting sensitive data in a log file. Data is logged into a file. The data in the log file is in a protected state and the data is never written to the log file in an unprotected fashion. Prior to the data being logged into the file, the data is parsed for specific data meeting predetermined criteria. The specific data is selectively protected with a security measure while leaving a remainder of the log file unprotected by the security measure. The viewer or program used to access the data in the log file is responsible for unprotecting or allowing the data to be viewed if the appropriate key is provided.

[0007]
Furthermore, U.S. Pat. No. 5,978,475 provides a method and an apparatus for generating a secure audit log using an untrusted machine communicating with a trusted machine over a limited communications channel. Entries are stored in the audit log in sequential order. Each entry in the audit log contains the oneway hash of the previous entry. This enables an auditor to verify that each entry was written into the log after the previous entry and before the subsequent entry. Any attempt to delete entries, add entries, or modify entries in the middle of the log will be immediately noticed because the oneway hash function values will no longer be valid.

[0008]
Each log entry contains a permission mask, an encrypted file, a (unkeyed) hash value including the encrypted file plus a representation of a previous hash value, and a (keyed) message authentication code (MAC) value that itself authenticates the hash value. The MAC is cryptographically secured with an authentication key derived by hashing an authentication key for a previous log entry; and the (encrypted file is cryptographically secured with an encryption key derived by hashing the authentication key. This makes it possible to give encryption keys for individual log entries to partiallytrusted entities, allowing them to read and decrypt files without being able to make undetectable changes. In addition, because both the authentication and encryption keys are sessionspecific and irreversibly related to their predecessor values (i.e., a current key can be generated from its predecessor, but not viceversa), an attack on a single entry can not be extended backward through the sequence to change the entire audit trail. This both prevents undetectable attack on a single entry and preserves the security of its predecessors.
OBJECT AND SUMMARY OF THE INVENTION

[0009]
The aim of the present invention is to provide a secure and reliable method for logging all the operations that occur in a complex environment for a subsequent control, linking these operations to the entities involved (namely, the system administrator and the system itself), and implementing privacy policies and security and antitampering functionalities.

[0010]
In particular, the aim of the present invention is to implement such policies and functionalities in such a way that:

 the log file does not immediately disclose its content;
 the log file cannot be modified without detection, i.e., if the log file is modified, this modification can be discovered a posteriori by the auditors that will check its content;
 the log entries can be decrypted and examined only by entities that have the rights to perform this operation; and,
 the log entries are not directly associated to the entities that are involved in the activities described in the log entries themselves.

[0015]
This aim is achieved by the present invention in that it relates to a method for protecting sensitive information, as defined in claims 1, to a processing system as claimed in claim 33, to a computer network as claimed in claim 34, and to a computer program product as defined in claim 35.

[0016]
Specifically, the need to keep the content of a log file unalterable (without detection) can be met by means of signatures and an hash chain that links all the log entries together, whereas the need to keep the content of a log file private is met by means of encryption. Each log entry is encrypted with a key that is successively encrypted with the keys of the entities that have the right to access the log entry. Moreover, the encrypted key may be distributed among a set of entities, if these entities have access to the log entry only together. The solution proposed also guarantees the privacy in the access to a log entry: this is obtained with an exclusion/elusion property according to which data to be used to access a log entry is encrypted in a way that it is impossible, if not in possession of the decrypting key, to decide if the data is useful or not for disclosing the content of the log entry. The consequence of this is that no one is able to decide whether an auditor has access or not to a log entry (except the auditor itself).

[0017]
This is efficient in the sense that it uses only one key to encrypt a log entry, and distributes this key among the auditors with the modes previously discussed. Nonetheless, this key changes for each log entry, leaving a fine granularity in giving or not the possibility to various auditors to access the content of a log entry.
BRIEF DESCRIPTION OF THE DRAWINGS

[0018]
For a better understanding of the present invention, a preferred embodiment, which is intended purely by way of example and is not to be construed as limiting, will now be described with reference to the attached drawings, wherein:

[0019]
FIG. 1 shows an untrusted computer where the present invention may be implemented;

[0020]
FIG. 2 shows a log window displaying some entries in a log file; and

[0021]
FIG. 3 shows a flow chart illustrating the operations carried out to encrypt a log entry according to the present invention.
DETAILED DESCRIPTION OF PREFERRED EMBODIMENTS OF THE INVENTION

[0022]
The following discussion is presented to enable a person skilled in the art to make and use the invention. Various modifications to the embodiments will be readily apparent to those skilled in the art, and the general principles herein may be applied to other embodiments and applications without departing from the spirit and scope of the present invention. Thus, the present invention is not intended to be limited to the embodiments shown, but is to be accorded the widest scope consistent with the principles and features disclosed herein and defined in the attached claims.

[0023]
Furthermore, the following discussion will refer to an environment in which administrators perform various activities on objects and these activities are logged in a file formed by a set of entries each referring to a particular event of interest in each activity.

[0024]
FIG. 1 shows a computer system 1 in which the present invention may be implemented and including a processing system 2 with a central processing unit (CPU), a video monitor 3, a keyboard 4, a mouse 5, storage devices 6 including floppy drives and other types of permanent and removable storage media, such as RAM, ROM and hard disk drives.

[0025]
FIG. 2 shows a log window displaying some entries in a log file. As shown in FIG. 2, each log entry is formed by two parts: an identification section that univocally identifies the entry and contains general information useful for retrieving and identifying the data, namely time stamps, administered entity and administrator, and a data section that contain the logged data which, in this case, represent the operations performed by a given administrator.

[0026]
FIG. 3 shows a flow chart illustrating the operations carried out to encrypt a log entry according to the present invention.

[0027]
In particular, in order to allow each log entry to be accessed by a single auditor or by a group of auditors separately one from another, data section in the log entry is encrypted by using a random key A_{i }which has a known fixed length, is different for each log entry, is generated automatically and randomly, and may be accessed by auditor(s) on the basis of privileges which define an exclusion/elusion property that will be discussed in detail later on (block 100). Then, for each auditor that is authorized to access the log entry, the random key A_{i }is encrypted and stored for later use in auditing. In particular, encryption of the random key A_{i }for each authorized auditor may be carried out in two different ways:

[0028]
a) each authorized auditor has his own symmetric secret key and the random key A_{i }is encrypted for each authorized auditor by using this symmetric secret key (block 110);

[0029]
b) each authorized auditor has his own pair of asymmetric public/private keys and the random key A_{i }is encrypted for each authorized auditor along with a known length random number R_{r }different for each authorized auditor by using the auditor's asymmetric public key (block 120). Preferably, before encryption, random key A_{i }and random number R_{r }are concatenated by arranging one after the other. The auditor will then use his private key to decrypt the encrypted aforementioned concatenation and to read out the known fixed length random key A_{i }so as to access the data section of the log entry. The need to combine the random key A_{i }with a random number R_{r }different for each authorized auditor derives from the need to preserve an elusion property which will be discussed later on.

[0030]
After encryption, the random key A_{i }is destroyed so that only the authorized auditor(s) will be able to reconstruct the original log entry (block 130).

[0031]
Furthermore, in order to achieve the exclusion/elusion property, the following approach is implemented which varies by using approach a) or b).

[0032]
In particular:

[0033]
c) in case approach a) has been implemented to encrypt the random key A_{i }for the authorized auditors, a fake key having the same properties as the random key A_{i }in terms of length, uniqueness and randomness is generated for the log entry and then encrypted for each auditor that is not authorized to access the log entry by using the auditor's symmetric secret key (block 140). The fake key for the log entry is obtained by computing H(A_{i}), i.e., a noninvertible function such as a oneway hash function of the random key A_{i}. In the literature many oneway hash functions easy to compute have been proposed (for example a oneway hash function is the one known as SHA1 (Secure Hash Algorithm, Revision 1) and described in National Institute of Standards and Technology, NIST FIPS PUB 1801, Secure Hash Standard, U.S. Department of Commerce, April 1995).

[0034]
Thanks to the property of the hash function, the fake key H(A_{i}) has the same properties as the random key A_{i }in terms of length, uniqueness and randomness and moreover given H(A_{i}), it is not feasible to compute A_{i}. The oneway hash function is an efficient source of randomness and is to be intended purely by way of example and not construed as limiting. In fact, a random key B_{i}, (which can be generated, for example, by a true hardware based random number generator or a pseudorandom generator passing FIPS 1402 test suite) different from the random key A_{i}, could also be directly generated and used as a fake key in place of H(A_{i}).

[0035]
d) in case approach b) has been implemented to encrypt the random key A_{i }for the authorized auditors, a fake key having the same properties as the random key A_{i }in terms of length, uniqueness and randomness is generated for the log entry and then encrypted for each auditor that is not authorized to access the log entry, along with a known length random number R_{r}, different for each nonauthorized auditor, by using the auditor's asymmetric public key (block 150). Similarly to approach c), the fake key for the log entry can be obtained by computing H(A_{i}), but a random key B_{i }different from the random key A_{i }could also be directly generated in place of H(A_{i}). The need to combine the fake key H(A_{i}) with a random number R_{r }different for each nonauthorized auditor, derives from the need to preserve an elusion property which will be discussed hereinafter.

[0036]
Then, an encrypted log entry is formed including the encrypted data section, the encrypted random keys A_{i }for the authorized auditors, the encrypted fake keys (e.g. H(A_{i})) for the nonauthorized auditors, and further data, as described in more detail later on (block 160).

[0037]
The preceding approach allows simultaneous achievement of the following two security properties:

 Exclusion: it is easy to exclude one or more auditors from accessing the encrypted data section of the log entry by simply providing them with the fake key. Advantageously, by using different random keys A_{i }for different log entries allows achievement of a fine granularity in giving access to each log entry. Thus the exclusion can be local to each log entry.

[0039]
Elusion: it may be appreciated that by simply looking at the encrypted log file it is not possible to understand which auditors have access to which log entries. This is due to the fact that a random key (A_{i}, H(A_{i})) is encrypted for each auditor. In this connection, a fine distinction between the two previously described approaches a) and b) can be made:

 approach a): access to an encrypted log entry depends on the possession of the random key A_{i}, that is necessary to decrypt the encrypted log entry, or the fake key H(A_{i}), that does not allow access to the encrypted log entry. However, as a good symmetric key encryption algorithm (e.g. AES) is supposed to make computationally unfeasible to understand if a certain encrypted text has been obtained from a specified clear text, (see also http://csrc.nist.gov/CryptoToolkit/aes/preround1/aes_{—}9709.htm section 4 as available on the Internet on 5 Apr., 2004, subsection SECURITY) it is unfeasible that a person who does not know the symmetric secret key for an auditor can deduce which key (whether A_{i }or H(A_{i})) has been encrypted for that auditor. It may be appreciated that it is important that the fake key H(A_{i}) changes for each log entry. In fact, if a constant value for the auditors that do not have access to a log entry is used, then encrypting a constant value by using a fixed key (the secret key of the nonauthorized auditor) will disclose the log entries that are not accessible to an auditor by simply inspecting the log file and looking for a repeated value for an auditor;
 approach b): because an asymmetric cryptosystem assume that secrecy of the public key is not a requirement it is possible to deduce which auditor is able to decrypt the encrypted random key A_{i}. Simply encrypting A_{i }with the possible auditor's public and checking which result matches the encrypted random key A_{i }retrieved from the log entry. In order to prevent such a disclosure of information, the use of random values R_{r }ensures the elusion property. In particular, for those auditors having the right to access the encrypted log entry, then the random key A_{i }is encrypted along with a random number R_{r }which is different for each auditor in order to ensure that an auditor decrypting the encrypted random key A_{i }is not able, through asymmetric encryption using the other's auditors public key, to deduce which of them has access to the log entry. In fact, without the random value, an auditor that knows A_{i }could try to encrypt A_{i }by using the known public keys of all the other auditors and then identify those auditors who have an autonomous access and those who don't. At the same time, for those auditors who do not have access to a log entry, a random value (also in this case, different for each auditor) is encrypted along with the fake key by using the public key of each auditor, thus the resulting value is undistinguishable from the encryption of the random key A_{i }and a random value.

[0042]
Here below are the structures of an encrypted log entry in the two previous approaches a) and b), where i indexes the log entries, k indexes the entities involved in the logged activity, and j indexes the various auditors:
L _{i} ={TS _{i} ,U _{h},
,
_{i} ,SE(
A _{i} /D _{i}),
SE(
K _{1} /A _{i}), . . . ,
SE(
K _{j1} /A _{i}),
SE(
K _{j} /H(
A _{i})), . . . ,
SE(
K _{n} /H(
A _{i})),
HC _{i} ,S _{i}} a>
L _{i} ={TS _{i} ,U _{k},
,
_{i} ,SE(
A _{i} /D _{i}),
AE(
K ^{+} _{1}/(
A _{i} ,R _{1})), . . . ,
AE(
K ^{+} _{j1}/(
A _{i} ,R _{j1})),
AE(
K ^{+} _{j} /H(
A _{i}),
R _{j}), . . . ,
AE(
K ^{+} _{j1}/(
H(
A _{i}),
R _{n})
HC _{i} ,S _{i}} b>
where:

 TS_{i }is the timestamp assigned to the entry. It may express the time of logging or the time of the reception. Even if the data contained in the log entry already contains a timestamp, TS_{i }may be useful for some cross checks on the data;
 U_{k }is an identifier of the log entry. It contains all the information necessary to identify the activity and the entities involved (administrator and administered system);
 _{i }is the length of data in cryptographic blocks;
 D_{i }are the logged data in the ith log entry;
 A_{i }is the random key used to encrypt the data in the ith log entry;
 K_{1}, . . . K_{n }are the auditors' symmetric secret keys used in approach a) to encrypt random key A_{i};
 K^{+} _{1 }. . . K^{+} _{n }are the auditors' asymmetric public keys used in approach b) to encrypt random key A_{i};
 R_{1}, . . . R_{n }are the random values used in approach b) to preserve the elusion property.
 SE(x/y) is a symmetric encryption function that uses the symmetric key x to encrypt data y and returns the encrypted data. This symmetric encryption function may be the one known as AES and described in National Institute of Standards and Technologies, NIST FIPS PUB 197, Advanced Encryption Standard (AES), U.S. Department of Commerce, November 2001;
 AE(x^{+}/y) is an asymmetric encryption function that uses the asymmetric key x^{+} to encrypt data y and returns the encrypted data. This asymmetric encryption function may be the one known as RSA and described in R. Rivest, A. Shamir, and L. Adleman, A Method for Obtaining Digital Signatures and PublicKey Cryptosystems, Communications of the ACM, 21 (1978), pp. 120126;
 H(x) is the oneway hash function, for example the one known as SHA1.
 HC_{i }is the link of the hash chain for the ith log entry;
 S_{i }is the signature of the link of the hash chain, i.e., it corresponds to Sign (B^{−}/HC_{i}), that is the function of digital signing HC_{i }with the logging system private key (B^{−}); it outputs the signature. Functions that may be used are, for example, the abovementioned RSA or the one known as DSA and described in National Institute of Standards and Technologies, NIST FIPS PUB 186, Digital Signature Standard, U.S. Department of Commerce, May 1994.

[0056]
In the previous two examples a) and b) SE(A_{i}/D_{i}) represents the encryption of the data D_{i }logged in the ith log line carried out by using the random key A_{i }in order to allow the access to the encrypted data only to the authorized auditors, whereas SE(K_{i}/H(A_{i})) represents the encryption of the fake key H(A_{i}) carried out by using the jth auditor's symmetric secret key K_{j}, and AE(K^{+} _{j}/(A_{i},R_{j})) represents the encryption of the random key A_{i }concatenated with the random value R_{j }carried out by using the jth auditor's public key K^{+} _{j}.

[0057]
It may therefore be appreciated that in previous two examples, auditors from 0 to j−1 have access to the log line content, whereas auditors from j to n have not.

[0058]
The element (link) HC
_{i }of the hash chain is the hash of the previous log entry hash (i.e. HC
_{i1}) concatenated with all the elements of the current entry, except HC
_{i }and S
_{i }(because the first one is what we are computing, and the second one will depend on the first one). The element HC
_{i }of the hash chain is computed by using the following formulas (for both the abovedescribed approaches a) and b)):
HC _{i} =H(
HC _{i1} ,TS _{i} ,U _{k},
,
_{i} ,SE(
A _{i} /D _{i}),
SE(
K _{1} /A _{i}), . . . ,
SE(
K _{j1} /A _{i}),
SE(
K _{j} /H(
A _{i})), . . . ,
SE(
K _{n} /H(
A _{i})) a>
HC _{i} =H(
HC _{i1} ,TS _{i} ,U _{k},
,
_{i} ,SE(
A _{i} /D _{i}),
AE(
K ^{+} _{1}/(
A _{i} ,R _{1})), . . . ,
AE(
K ^{+} _{j1}/(
A _{i} ,R _{j1})),
AE(
K ^{+} _{j} /H(
A _{i}),
R _{j}), . . . ,
AE(
K ^{+} _{n} /H(
A _{i}),
R _{n})) b>

[0059]
The first element of the hash chain, namely HC_{1}, is computed using as previous element a fixed and known value for HC_{0 }which may be recorded, without encryption, in the beginning of the log file. When a line is verified, the hash of the previous line is trusted, thus a verification of the signature of the previous line is performed.

[0060]
In addition to linking log entries, hash chain with signed elements makes it possible to control the integrity of the chain even if the log file is stored (also partly) in different removable memory devices, on condition that the first element of the chain be authentic.

[0061]
Each element of the hash chain is singularly signed by using the logging system private key B^{−} because anyone could compute an hash without knowing any secret and therefore anyone could modify, delete or invert log entries without possibility of detection: it may in fact be possible to a malicious user to recompute the elements of the hash chain subsequent to the one in the maliciously modified log entry.

[0062]
To detect a possible corruption of the log file, it is not enough to verify authenticity of the last chain element only, but it is necessary to verify authenticity of all the chain elements, i.e., it is necessary to recompute the whole hash chain, compare it with the existing one and verify authenticity of every single signature for every single chain element. In fact, in case a malicious user modifies the last log entry and update the corresponding chain element, such a user does not know the logging system private key B^{−} and therefore cannot sign the modified chain element. During verification of all the chain elements, including those ones generated after log file corruption, the chain elements would however prove to be linked as expected and the signature of the last chain element would prove to be authentic.

[0063]
As an alternative to the abovedescribed asymmetric signature function, a function allowing computation of a MAC (Message Authentication Code) could be used, which is a function that uses a secret altogether similar to a symmetric key to generate an authentic digest. By even using MAC functions based on hash, i.e., HMAC functions, it is also possible to compute an HMAC chain which is intrinsically and automatically authentic.

[0064]
Heretofore there have been discussed the socalled “single auditing”, i.e., the possibility of enabling an auditor to access a log entry, and the socalled “separate auditing”, i.e., the possibility of enabling a group of auditors to access a log entry separately one from another.

[0065]
Hereinafter there will be discussed the socalled “group auditing”, i.e., the possibility of enabling a group of auditors to access a log entry only when at least any n authorized auditors over N in the group agree on looking at its content.

[0066]
This functionality is based on what in literature is known as “secret sharing”.

[0067]
For example, G. R. Blakley, Safeguarding cryptographic keys, Proc. of AFIPS, 1979 NCC, Vol. 48, Arlington, Va., June 1979, pp. 313317 proposes a model for sharing a secret based on ndimensional geometry and in particular develops a method to keep copies of cryptographic keys by distributing the secret among many people. An example of this method may also be found in A. J. Menezes, P. C. van Oorschot and S. A. Vanstone, Handbook of Applied Cryptography, CRC Press, 1996, pags 524 to 528 which discloses distributing a secret among m parts, where at least three of them are required to reconstruct the secret. In this paper, the secret is represented as a point into the three dimensional space and m planes are constructed such that any three of them intersect into the point representing the secret, but any two of them define a line. Only knowing at least three of these planes, it is possible to unambiguously identify the single point in space.

[0068]
Another proposal may be found in A. Shamir, How to share a secret, Communications of the ACM, 22 (1979), pp. 612613, which discloses a method for distributing a secret among n entities where at least k of them are necessary to reconstruct the original secret. It is based on polynomial interpolation. The idea is to construct a polynomial q(x) of degree k−1 having random coefficients, except for the coefficient of x_{0}, that is equal to the secret to share. Distributing the evaluation of the polynomial into n different points, then it is possible to calculate the k coefficients (thus the shared secret also) only when at least k evaluations of the polynomial are available. This can be done by interpolating the polynomial in the k points (note that there is only one polynomial of degree k−1 that fits into k points, and there are infinite polynomials of the same degree that fit into k−1 points).

[0069]
The group auditing of the present invention can be based on the method disclosed in the latter paper and aims at achieving the following functionalities:

 each auditor can access the content of a log line either alone (if he has the rights) or with the cooperation of other auditors (if he belongs to a group of auditors that can have access to the line);
 when a group of auditors has used a secret to disclose the content of a line, then this secret is useless if used to disclose the content of other lines; the reason is that when a group of auditors agree on looking at the content of a line, then some of them may not agree on disclosing the content of other lines to the members of the same group;
 each auditor may belong to any number of groups or to none.

[0073]
To achieve the aforementioned functionalities, a share to determine the random key A_{i }is distributed among the auditors that need a group access to a log line. That is, in block 110, instead of encrypting the random key A_{i}, data that allows the reconstruction of the complete random key A_{i }when combined with complementary data owned by other authorized auditors is encrypted for each authorized auditor.

[0074]
This implies that to decrypt a log entry there may be:

 auditors that have access to the log entry as stand alone entities, i.e. they have SE(K_{j}/A_{i}) or AE(K^{+} _{j}/(A_{i},R_{j}));
 auditors that do not have access to the log entry as stand alone entities, i.e. they have SE(K_{j}/H(A_{i})) in approach a) or AE(K^{+} _{j}/(H(A_{i}),R_{j})) in approach b);
 auditors that have access to the log entry only with the collaboration of at least k auditors, i.e. they have SE(K_{j}/ΣA_{i}) in approach a) or AE(K^{+} _{j}/ΣA_{i},R_{j}) in approach b), where ΣA_{i }is the share of a random key A_{i }that allows reconstruction of the random key A_{i }with the collaboration of other k−1 auditors. Auditors may belong to many groups, thus having many shares of the secret (obviously, the various shares will be related to different polynomials).

[0078]
Note that the previously listed three sets of auditors may be not disjoint (the first two are disjoint). Thus, the invention allows for auditors that may access a log entry by themselves, or in collaboration with other auditors also, or only when other auditors in the group agree on disclosing the content of a log entry.

[0079]
To add an auditor to a group, it is sufficient to give the auditor a new share based on the polynomial, encrypting this share with the auditor's key. To exclude an auditor from a group it is sufficient not to give the auditor his share anymore.

[0080]
To modify the minimum number of auditors necessary to decrypt a log entry, a different polynomial is used, as thought in the abovementioned How to share a secret.

[0081]
To work properly and to be able to decrypt correctly a log entry for a group, at least the following information for each group are required:

 a group identifier (ID group);
 the minimum number of auditors necessary to disclose the secret;
 the identifiers of all the auditors belonging to the group.

[0085]
For each auditor that potentially has access to a log entry the following data are stored:

 approach a):
SE(K_{j}/[H(A_{i}), ID_{group′}, Σ′A_{i}, ID_{group″}, Σ″A_{i}, . . . ])
 approach b):
AE(K^{+} _{j}/[H(A_{i}), R_{j}, ID_{group′}, Σ′A_{i}, ID_{group″}, Σ″A_{i}, . . . ]R_{j})

[0088]
In this example the jth auditor has not access to the log entry as individual, but only as belonging to some groups. If an auditor does not belong to a group (or a group does not have access to the log entry) then Σ may be left as a set of zeroes.

[0089]
Note that it is not necessary to encrypt fake shares for the nonauthorized auditors to preserve elusion property because unpredictability of the encrypted data is already provided by encryption of A_{i }or H(A_{i}).

[0090]
Group auditing applied to multiple groups does not jeopardize the discussed security policies, in particular even if shares of different groups on the random key A_{i }are joined together, this does not allow the determination of the random key A_{i}.

[0091]
To demonstrate this, let's suppose the worst case, i.e. let's imagine the case in which m′−1 auditors of a group (requiring m′ auditors to compute A_{i}) colluding with m″−1 auditors of another group (requiring m″ auditors to compute A_{i}), the two groups may overlap or not.

[0092]
The two polynomials that it is necessary to determine are:
y=α _{m′1} x ^{m′−1}+α_{m′2} x ^{m′−2}+ . . . +α_{1} x+A _{i }
y=β _{m″1} x ^{m″−1}+β_{m″2} x ^{m″−2}+ . . . +β_{1} x+A _{i }

[0093]
The target is to determine α and β values and A_{i}, i.e., m′+m″−1 values in all. The colluding auditors have m′+m″−2 points (possibly not distinct), m′−1 from one polynomial, and m″−1 from the other polynomial. This allows to write a system of m′+m″−2 equations with m′+m″−1 variables. The target may not be reached because the system of equations is undetermined if the assumption is made that a single polynomial of degree m is undetermined if only m−1 points are available. However, to discover the shared key, it is sufficient to determine A_{i}, but this is not possible. Let's denote with P the set of equations coming from the first polynomial and with Q the set of equations coming from the second polynomial.

[0094]
Given that A_{i }cannot be determined from P (see the abovementioned How to share a secret), then reduction of this set leads to an equation of this kind:
c _{1}α_{j} +c _{2} A _{i} =b _{1 }
For the same reason, reduction of Q leads to:
C _{3}β_{k} +c _{4} A _{i} =b _{2 }
where the c_{m }and b_{n }are constant values.

[0095]
The system of these two equations does not allow to determine A_{i }because a_{j }and b_{k }are different unknowns (they are coefficients from different polynomials). Therefore, even if different auditors from different groups collude to determine the shared key, they will not be able to get it unless the required number of auditors in one of the groups is reached.

[0096]
The same demonstration applies to the case where two auditors belonging to different groups own the same share (i.e. the same point in the plane, where two distinct polynomials intersect).

[0097]
In addition, to the previously discussed “group auditing”, the socalled “category auditing” may also be provided, to meet a need that may arise in some contexts in which one or more auditors belonging to a given category can cooperate with one or more auditors belonging to a different category or different categories.

[0098]
This need may be met by applying the previously discussed secret sharing algorithm to m categories, thus generating m subsecrets, and then by applying again the secret sharing algorithm to the m subsecrets, imposing at the same time that k elements over n in each category are necessary to reconstruct the respective subsecret, where k is lower than or equal to n and that k and n may change for each category.

[0099]
Finally, various modifications to the embodiments will be readily apparent to those skilled in the art, and the generic principles herein may be applied to other embodiments and applications without departing from the spirit and scope of the present invention, as defined by the appended claims.

[0100]
For example, instead of generating a fake key for a log entry having the same properties as the random key A_{i }in terms of length, uniqueness and randomness, and then encrypting it by using the nonauthorized auditors' secret or public keys, elusion property may also be preserved by directly generating a fake key for the log entry having the same properties as the encrypted random key E(K/A_{i}) in terms of length, uniqueness and randomness. Such a fake key may for example be generated by computing a hash on the encrypted random key E(K/A_{i}).

[0101]
Furthermore, instead of encrypting a fake key for a log entry by using the nonauthorized auditors' keys, the fake key may also be encrypted by using randomly generated keys different for each nonauthorized auditor.

[0102]
In addition, the preceding description discloses the present invention in relation to sensitive information represented by random keys A_{i }used to encrypt data sections in the log entries. It may be appreciated that the present invention may also be applied to sensitive information represented directly by data sections in the log entries.

[0103]
This embodiment differs from the one previously discussed only in the item of sensitive information subjected to encryption: in the first embodiment this item of sensitive information is q random key A_{i}, whereas in this second embodiment this item of sensitive information is data section in a log entry. Therefore, for each auditor that is authorized to access a log entry, in approach a) data section D_{i }is encrypted by using the auditor's secret key K_{j}, whereas in approach b) data section D_{i }is encrypted along with a random value R_{r }different for each authorized auditor by using the auditor's public key K^{+} _{j}. Instead, for each auditor that is not authorized to access the log entry, in approach a) a fake data section generated by computing a hash on the data section D_{i }or randomly generated is encrypted by using the auditor's secret key K_{j}, whereas in approach b) a fake data section generated by computing a hash on the data section D_{i }or randomly generated is encrypted along with a random value R_{r }different for each nonauthorized auditor by using the auditor's public key K^{+} _{j}.

[0104]
Group and category auditing may also be applied to this embodiment by distributing shares ΣD_{i }of the data section D_{i }among the various categories and the auditors in the categories.

[0105]
In general, for each item of sensitive information to be protected (e.g., for each random key A_{i }or for each data section D_{i}) an item of fake information may be generated so that it is not possible or practicable for an expert to envisage a test for distinguishing the item of fake information from the item of sensitive information, i.e., for understanding which one is the item of sensitive information and which one is the item of fake information.

[0106]
For example, the above criteria of having the same properties in terms of length may imply that the same length is adopted for the item of fake information as for the item of sensitive information to be protected. Another possibility is that different lengths are adopted for different items of sensitive information (e.g., within a predetermined range of lengths) so that it will not be possible to distinguish between items of fake and sensitive information based on the length of the item of sensitive information.

[0107]
As to the uniqueness of an item of fake information, this is to be intended in a statistical sense, i.e., each item of fake information should be generated so that it is highly likely that it is different from another item of fake information (and from the item of sensitive information to be protected), although a coincidence of items of fake information, although highly unlikely, remains possible.

[0108]
A criterion for establishing randomness of an item of fake information with respect to an item of sensitive information to be protected is, for example, that the item of fake information is generated by a true hardwarebased random number generator or by a pseudorandom generator passing FIPS 1402 test suite.